Mingyang Zhang

Cheng Li, Mingyang Zhang, Qiaozhu Mei et al. · baidu, deepmind

Personalized text generation is an emerging research area that has attracted much attention in recent years. Most studies in this direction focus on a particular domain by designing bespoke features or models. In this work, we propose a general approach for personalized text generation using large language models (LLMs). Inspired by the practice of writing education, we develop a multistage and multitask framework to teach LLMs for personalized generation. In writing instruction, the task of writing from sources is often decomposed into multiple steps that involve finding, evaluating, summarizing, synthesizing, and integrating information. Analogously, our approach to personalized text generation consists of multiple stages: retrieval, ranking, summarization, synthesis, and generation. In addition, we introduce a multitask setting that helps the model improve its generation ability further, which is inspired by the observation in education that a student's reading proficiency and writing ability are often correlated. We evaluate our approach on three public datasets, each of which covers a different and representative domain. Our results show significant improvements over a variety of baselines.

Changlong Jiang, Yang Xiao, Cunlin Wu et al.

3D interacting hand pose estimation from a single RGB image is a challenging task, due to serious self-occlusion and inter-occlusion towards hands, confusing similar appearance patterns between 2 hands, ill-posed joint position mapping from 2D to 3D, etc.. To address these, we propose to extend A2J-the state-of-the-art depth-based 3D single hand pose estimation method-to RGB domain under interacting hand condition. Our key idea is to equip A2J with strong local-global aware ability to well capture interacting hands' local fine details and global articulated clues among joints jointly. To this end, A2J is evolved under Transformer's non-local encoding-decoding framework to build A2J-Transformer. It holds 3 main advantages over A2J. First, self-attention across local anchor points is built to make them global spatial context aware to better capture joints' articulation clues for resisting occlusion. Secondly, each anchor point is regarded as learnable query with adaptive feature learning for facilitating pattern fitting capacity, instead of having the same local representation with the others. Last but not least, anchor point locates in 3D space instead of 2D as in A2J, to leverage 3D pose prediction. Experiments on challenging InterHand 2.6M demonstrate that, A2J-Transformer can achieve state-of-the-art model-free performance (3.38mm MPJPE advancement in 2-hand case) and can also be applied to depth domain with strong generalization.

Mingyang Zhang, Xinyi Yu, Haodong Zhao et al.

One-shot Neural architecture search (One-shot NAS) has been proposed as a time-efficient approach to obtain optimal subnet architectures and weights under different complexity cases by training only once. However, the subnet performance obtained by weight sharing is often inferior to the performance achieved by retraining. In this paper, we investigate the performance gap and attribute it to the use of uniform sampling, which is a common approach in supernet training. Uniform sampling concentrates training resources on subnets with intermediate computational resources, which are sampled with high probability. However, subnets with different complexity regions require different optimal training strategies for optimal performance. To address the problem of uniform sampling, we propose ShiftNAS, a method that can adjust the sampling probability based on the complexity of subnets. We achieve this by evaluating the performance variation of subnets with different complexity and designing an architecture generator that can accurately and efficiently provide subnets with the desired complexity. Both the sampling probability and the architecture generator can be trained end-to-end in a gradient-based manner. With ShiftNAS, we can directly obtain the optimal model architecture and parameters for a given computational complexity. We evaluate our approach on multiple visual network models, including convolutional neural networks (CNNs) and vision transformers (ViTs), and demonstrate that ShiftNAS is model-agnostic. Experimental results on ImageNet show that ShiftNAS can improve the performance of one-shot NAS without additional consumption. Source codes are available at https://github.com/bestfleer/ShiftNAS.

Junchen Lu, Berrak Sisman, Mingyang Zhang et al.

The goal of Automatic Voice Over (AVO) is to generate speech in sync with a silent video given its text script. Recent AVO frameworks built upon text-to-speech synthesis (TTS) have shown impressive results. However, the current AVO learning objective of acoustic feature reconstruction brings in indirect supervision for inter-modal alignment learning, thus limiting the synchronization performance and synthetic speech quality. To this end, we propose a novel AVO method leveraging the learning objective of self-supervised discrete speech unit prediction, which not only provides more direct supervision for the alignment learning, but also alleviates the mismatch between the text-video context and acoustic features. Experimental results show that our proposed method achieves remarkable lip-speech synchronization and high speech quality by outperforming baselines in both objective and subjective evaluations. Code and speech samples are publicly available.

Cheng Li, Mingyang Zhang, Qiaozhu Mei et al.

Facilitated by large language models (LLMs), personalized text generation has become a rapidly growing research direction. Most existing studies focus on designing specialized models for a particular domain, or they require fine-tuning the LLMs to generate personalized text. We consider a typical scenario in which the large language model, which generates personalized output, is frozen and can only be accessed through APIs. Under this constraint, all one can do is to improve the input text (i.e., text prompts) sent to the LLM, a procedure that is usually done manually. In this paper, we propose a novel method to automatically revise prompts for personalized text generation. The proposed method takes the initial prompts generated by a state-of-the-art, multistage framework for personalized generation and rewrites a few critical components that summarize and synthesize the personal context. The prompt rewriter employs a training paradigm that chains together supervised learning (SL) and reinforcement learning (RL), where SL reduces the search space of RL and RL facilitates end-to-end training of the rewriter. Using datasets from three representative domains, we demonstrate that the rewritten prompts outperform both the original prompts and the prompts optimized via supervised learning or reinforcement learning alone. In-depth analysis of the rewritten prompts shows that they are not only human readable, but also able to guide manual revision of prompts when there is limited resource to employ reinforcement learning to train the prompt rewriter, or when it is costly to deploy an automatic prompt rewriter for inference.

Mingyang Zhang, Ziqi Di, Maoguo Gong et al.

In recent years, research on hyperspectral image (HSI) classification has continuous progress on introducing deep network models, and recently the graph convolutional network (GCN) based models have shown impressive performance. However, these deep learning frameworks based on point estimation suffer from low generalization and inability to quantify the classification results uncertainty. On the other hand, simply applying the Bayesian Neural Network (BNN) based on distribution estimation to classify the HSI is unable to achieve high classification accuracy due to the large amount of parameters. In this paper, we design a Bayesian layer with Bayesian idea as an insertion layer into point estimation based neural networks, and propose a Bayesian Layer Graph Convolutional Network (BLGCN) model by combining graph convolution operations, which can effectively extract graph information and estimate the uncertainty of classification results. Moreover, a Generative Adversarial Network (GAN) is built to solve the sample imbalance problem of HSI dataset. Finally, we design a dynamic control training strategy based on the confidence interval of the classification results, which will terminate the training early when the confidence interval reaches the preseted threshold. The experimental results show that our model achieves a balance between high classification accuracy and strong generalization. In addition, it can quantifies the uncertainty of the classification results.

Xiaolong Fan, Maoguo Gong, Yue Wu et al.

The empirical studies of Graph Neural Networks (GNNs) broadly take the original node feature and adjacency relationship as singleview input, ignoring the rich information of multiple graph views. To circumvent this issue, the multiview graph analysis framework has been developed to fuse graph information across views. How to model and integrate shared (i.e. consistency) and view-specific (i.e. complementarity) information is a key issue in multiview graph analysis. In this paper, we propose a novel Multiview Variational Graph Information Bottleneck (MVGIB) principle to maximize the agreement for common representations and the disagreement for view-specific representations. Under this principle, we formulate the common and view-specific information bottleneck objectives across multiviews by using constraints from mutual information. However, these objectives are hard to directly optimize since the mutual information is computationally intractable. To tackle this challenge, we derive variational lower and upper bounds of mutual information terms, and then instead optimize variational bounds to find the approximate solutions for the information objectives. Extensive experiments on graph benchmark datasets demonstrate the superior effectiveness of the proposed method.

Yaqing Wang, Jiepu Jiang, Mingyang Zhang et al.

Personalized text generation presents a specialized mechanism for delivering content that is specific to a user's personal context. While the research progress in this area has been rapid, evaluation still presents a challenge. Traditional automated metrics such as BLEU and ROUGE primarily measure lexical similarity to human-written references, and are not able to distinguish personalization from other subtle semantic aspects, thus falling short of capturing the nuances of personalized generated content quality. On the other hand, human judgments are costly to obtain, especially in the realm of personalized evaluation. Inspired by these challenges, we explore the use of large language models (LLMs) for evaluating personalized text generation, and examine their ability to understand nuanced user context. We present AuPEL, a novel evaluation method that distills three major semantic aspects of the generated text: personalization, quality and relevance, and automatically measures these aspects. To validate the effectiveness of AuPEL, we design carefully controlled experiments and compare the accuracy of the evaluation judgments made by LLMs versus that of judgements made by human annotators, and conduct rigorous analyses of the consistency and sensitivity of the proposed metric. We find that, compared to existing evaluation metrics, AuPEL not only distinguishes and ranks models based on their personalization abilities more accurately, but also presents commendable consistency and efficiency for this task. Our work suggests that using LLMs as the evaluators of personalized text generation is superior to traditional text similarity metrics, even though interesting new challenges still remain.

Ruofan Wu, Mingyang Zhang, Lingjuan Lyu et al.

The paradigm of vertical federated learning (VFL), where institutions collaboratively train machine learning models via combining each other's local feature or label information, has achieved great success in applications to financial risk management (FRM). The surging developments of graph representation learning (GRL) have opened up new opportunities for FRM applications under FL via efficiently utilizing the graph-structured data generated from underlying transaction networks. Meanwhile, transaction information is often considered highly sensitive. To prevent data leakage during training, it is critical to develop FL protocols with formal privacy guarantees. In this paper, we present an end-to-end GRL framework in the VFL setting called VESPER, which is built upon a general privatization scheme termed perturbed message passing (PMP) that allows the privatization of many popular graph neural architectures.Based on PMP, we discuss the strengths and weaknesses of specific design choices of concrete graph neural architectures and provide solutions and improvements for both dense and sparse graphs. Extensive empirical evaluations over both public datasets and an industry dataset demonstrate that VESPER is capable of training high-performance GNN models over both sparse and dense graphs under reasonable privacy budgets.

Hao Liu, Yunhao Gao, Wei Li et al.

Deep learning-based methods have achieved significant success in remote sensing Earth observation data analysis. Numerous feature fusion techniques address multimodal remote sensing image classification by integrating global and local features. However, these techniques often struggle to extract structural and detail features from heterogeneous and redundant multimodal images. With the goal of introducing frequency domain learning to model key and sparse detail features, this paper introduces the spatial-spectral-frequency interaction network (S$^2$Fin), which integrates pairwise fusion modules across the spatial, spectral, and frequency domains. Specifically, we propose a high-frequency sparse enhancement transformer that employs sparse spatial-spectral attention to optimize the parameters of the high-frequency filter. Subsequently, a two-level spatial-frequency fusion strategy is introduced, comprising an adaptive frequency channel module that fuses low-frequency structures with enhanced high-frequency details, and a high-frequency resonance mask that emphasizes sharp edges via phase similarity. In addition, a spatial-spectral attention fusion module further enhances feature extraction at intermediate layers of the network. Experiments on four benchmark multimodal datasets with limited labeled data demonstrate that S$^2$Fin performs superior classification, outperforming state-of-the-art methods. The code is available at https://github.com/HaoLiu-XDU/SSFin.

Mingyang Zhang, Hao Chen, Chunhua Shen et al.

Large Language Models (LLMs), such as LLaMA and T5, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LLMs on downstream tasks, their deployment is still hindered by the vast model scale and computational costs. Post-training model pruning offers a way to compress LLMs. However, the current pruning methods designed for LLMs are not compatible with LoRA. This is due to their utilization of unstructured pruning on LLMs, impeding the merging of LoRA weights, or their dependence on the gradients of pre-trained weights to guide pruning, which can impose significant memory overhead. To this end, we propose LoRAPrune, a new framework that delivers an accurate structured pruned model in a highly memory-efficient manner. Specifically, we first design a LoRA-guided pruning criterion, which uses the weights and gradients of LoRA, rather than the gradients of pre-trained weights for importance estimation. We subsequently integrate this criterion into an iterative pruning process, effectively removing redundant channels and heads. Extensive experimental results demonstrate the superior performance of our LoRAPrune over existing approaches on the LLaMA series models. At a 50\% compression rate, LoRAPrune demonstrates superior performance over LLM-Pruner, achieving a reduction in perplexity by 4.81 on WikiText2 and 3.46 on PTB, while also decreasing memory usage by 52.6%. Besides, LoRAPrune also matches semi-structural pruning across multiple LLMs, proving its wide applicability. The code is available at https://github.com/aim-uofa/LoRAPrune.

Liu Yang, Mingyang Zhang, Cheng Li et al.

Many natural language processing and information retrieval problems can be formalized as the task of semantic matching. Existing work in this area has been largely focused on matching between short texts (e.g., question answering), or between a short and a long text (e.g., ad-hoc retrieval). Semantic matching between long-form documents, which has many important applications like news recommendation, related article recommendation and document clustering, is relatively less explored and needs more research effort. In recent years, self-attention based models like Transformers and BERT have achieved state-of-the-art performance in the task of text matching. These models, however, are still limited to short text like a few sentences or one paragraph due to the quadratic computational complexity of self-attention with respect to input text length. In this paper, we address the issue by proposing the Siamese Multi-depth Transformer-based Hierarchical (SMITH) Encoder for long-form document matching. Our model contains several innovations to adapt self-attention models for longer text input. In order to better capture sentence level semantic relations within a document, we pre-train the model with a novel masked sentence block language modeling task in addition to the masked word language modeling task used by BERT. Our experimental results on several benchmark datasets for long-form document matching show that our proposed SMITH model outperforms the previous state-of-the-art models including hierarchical attention, multi-depth attention-based hierarchical recurrent neural network, and BERT. Comparing to BERT based baselines, our model is able to increase maximum input text length from 512 to 2048. We will open source a Wikipedia based benchmark dataset, code and a pre-trained checkpoint to accelerate future research on long-form document matching.

Zixuan Ke, Weize Kong, Cheng Li et al.

Large Language Models (LLMs) have demonstrated superior results across a wide range of tasks, and Retrieval-augmented Generation (RAG) is an effective way to enhance the performance by locating relevant information and placing it into the context window of the LLM. However, the relationship between retrievers and LLMs in a RAG is still under-investigated. Most existing work treats the retriever and the LLM as independent components and leaves a gap between retrieving human-"friendly" information and assembling a LLM-"friendly" context. In this work, we examine a novel bridge mechanism. We validate the ranking and selection assumptions of retrievers in the context of RAG and propose a framework that chains together supervised and reinforcement learning to train a bridge model that optimizes the connection between the retriever and the LLM. Empirical results demonstrate the effectiveness of our method in both question-answering and personalized generation tasks.

Zhaoyang Wang, Dongyang Li, Mingyang Zhang et al.

Existing hyperspectral image (HSI) super-resolution (SR) methods struggle to effectively capture the complex spectral-spatial relationships and low-level details, while diffusion models represent a promising generative model known for their exceptional performance in modeling complex relations and learning high and low-level visual features. The direct application of diffusion models to HSI SR is hampered by challenges such as difficulties in model convergence and protracted inference time. In this work, we introduce a novel Group-Autoencoder (GAE) framework that synergistically combines with the diffusion model to construct a highly effective HSI SR model (DMGASR). Our proposed GAE framework encodes high-dimensional HSI data into low-dimensional latent space where the diffusion model works, thereby alleviating the difficulty of training the diffusion model while maintaining band correlation and considerably reducing inference time. Experimental results on both natural and remote sensing hyperspectral datasets demonstrate that the proposed method is superior to other state-of-the-art methods both visually and metrically.

Lei Yu, Meng Han, Yiming Li et al.

Vertical Federated Learning (VFL) is a federated learning paradigm where multiple participants, who share the same set of samples but hold different features, jointly train machine learning models. Although VFL enables collaborative machine learning without sharing raw data, it is still susceptible to various privacy threats. In this paper, we conduct the first comprehensive survey of the state-of-the-art in privacy attacks and defenses in VFL. We provide taxonomies for both attacks and defenses, based on their characterizations, and discuss open challenges and future research directions. Specifically, our discussion is structured around the model's life cycle, by delving into the privacy threats encountered during different stages of machine learning and their corresponding countermeasures. This survey not only serves as a resource for the research community but also offers clear guidance and actionable insights for practitioners to safeguard data privacy throughout the model's life cycle.

Zhaoyang Wang, Bo Hu, Mingyang Zhang et al.

Existing free-energy guided No-Reference Image Quality Assessment (NR-IQA) methods still suffer from finding a balance between learning feature information at the pixel level of the image and capturing high-level feature information and the efficient utilization of the obtained high-level feature information remains a challenge. As a novel class of state-of-the-art (SOTA) generative model, the diffusion model exhibits the capability to model intricate relationships, enabling a comprehensive understanding of images and possessing a better learning of both high-level and low-level visual features. In view of these, we pioneer the exploration of the diffusion model into the domain of NR-IQA. Firstly, we devise a new diffusion restoration network that leverages the produced enhanced image and noise-containing images, incorporating nonlinear features obtained during the denoising process of the diffusion model, as high-level visual information. Secondly, two visual evaluation branches are designed to comprehensively analyze the obtained high-level feature information. These include the visual compensation guidance branch, grounded in the transformer architecture and noise embedding strategy, and the visual difference analysis branch, built on the ResNet architecture and the residual transposed attention block. Extensive experiments are conducted on seven public NR-IQA datasets, and the results demonstrate that the proposed model outperforms SOTA methods for NR-IQA.

Ling Team, Bin Han, Caizhi Tang et al.

In this technical report, we present the Ring-linear model series, specifically including Ring-mini-linear-2.0 and Ring-flash-linear-2.0. Ring-mini-linear-2.0 comprises 16B parameters and 957M activations, while Ring-flash-linear-2.0 contains 104B parameters and 6.1B activations. Both models adopt a hybrid architecture that effectively integrates linear attention and softmax attention, significantly reducing I/O and computational overhead in long-context inference scenarios. Compared to a 32 billion parameter dense model, this series reduces inference cost to 1/10, and compared to the original Ring series, the cost is also reduced by over 50%. Furthermore, through systematic exploration of the ratio between different attention mechanisms in the hybrid architecture, we have identified the currently optimal model structure. Additionally, by leveraging our self-developed high-performance FP8 operator library-linghe, overall training efficiency has been improved by 50%. Benefiting from the high alignment between the training and inference engine operators, the models can undergo long-term, stable, and highly efficient optimization during the reinforcement learning phase, consistently maintaining SOTA performance across multiple challenging complex reasoning benchmarks.

Tao Chen, Mingyang Zhang, Jing Lu et al.

The pre-trained language model (eg, BERT) based deep retrieval models achieved superior performance over lexical retrieval models (eg, BM25) in many passage retrieval tasks. However, limited work has been done to generalize a deep retrieval model to other tasks and domains. In this work, we carefully select five datasets, including two in-domain datasets and three out-of-domain datasets with different levels of domain shift, and study the generalization of a deep model in a zero-shot setting. Our findings show that the performance of a deep retrieval model is significantly deteriorated when the target domain is very different from the source domain that the model was trained on. On the contrary, lexical models are more robust across domains. We thus propose a simple yet effective framework to integrate lexical and deep retrieval models. Our experiments demonstrate that these two models are complementary, even when the deep model is weaker in the out-of-domain setting. The hybrid model obtains an average of 20.4% relative gain over the deep retrieval model, and an average of 9.54% over the lexical model in three out-of-domain datasets.

Sergey Nikonorov, Berrak Sisman, Mingyang Zhang et al.

Conventional vocoders are commonly used as analysis tools to provide interpretable features for downstream tasks such as speech synthesis and voice conversion. They are built under certain assumptions about the signals following signal processing principle, therefore, not easily generalizable to different audio, for example, from speech to singing. In this paper, we propose a deep neural analyzer, denoted as DeepA - a neural vocoder that extracts F0 and timbre/aperiodicity encoding from the input speech that emulate those defined in conventional vocoders. Therefore, the resulting parameters are more interpretable than other latent neural representations. At the same time, as the deep neural analyzer is learnable, it is expected to be more accurate for signal reconstruction and manipulation, and generalizable from speech to singing. The proposed neural analyzer is built based on a variational autoencoder (VAE) architecture. We show that DeepA improves F0 estimation over the conventional vocoder (WORLD). To our best knowledge, this is the first study dedicated to the development of a neural framework for extracting learnable vocoder-like parameters.

Jingtao Rong, Xinyi Yu, Mingyang Zhang et al.

Adequate labeled data and expensive compute resources are the prerequisites for the success of neural architecture search(NAS). It is challenging to apply NAS in meta-learning scenarios with limited compute resources and data. In this paper, an across-task neural architecture search (AT-NAS) is proposed to address the problem through combining gradient-based meta-learning with EA-based NAS to learn over the distribution of tasks. The supernet is learned over an entire set of tasks by meta-learning its weights. Architecture encodes of subnets sampled from the supernet are iteratively adapted by evolutionary algorithms while simultaneously searching for a task-sensitive meta-network. Searched meta-network can be adapted to a novel task via a few learning steps and only costs a little search time. Empirical results show that AT-NAS surpasses the related approaches on few-shot classification accuracy. The performance of AT-NAS on classification benchmarks is comparable to that of models searched from scratch, by adapting the architecture in less than an hour from a 5-GPU-day pretrained meta-network.

Junchen Lu, Berrak Sisman, Rui Liu et al.

In this paper, we formulate a novel task to synthesize speech in sync with a silent pre-recorded video, denoted as automatic voice over (AVO). Unlike traditional speech synthesis, AVO seeks to generate not only human-sounding speech, but also perfect lip-speech synchronization. A natural solution to AVO is to condition the speech rendering on the temporal progression of lip sequence in the video. We propose a novel text-to-speech model that is conditioned on visual input, named VisualTTS, for accurate lip-speech synchronization. The proposed VisualTTS adopts two novel mechanisms that are 1) textual-visual attention, and 2) visual fusion strategy during acoustic decoding, which both contribute to forming accurate alignment between the input text content and lip motion in input lip sequence. Experimental results show that VisualTTS achieves accurate lip-speech synchronization and outperforms all baseline systems.

Mingyang Zhang, Xinyi Yu, Jingtao Rong et al.

In the past years, significant improvements in the field of neural architecture search(NAS) have been made. However, it is still challenging to search for efficient networks due to the gap between the searched constraint and real inference time exists. To search for a high-performance network with low inference time, several previous works set a computational complexity constraint for the search algorithm. However, many factors affect the speed of inference(e.g., FLOPs, MACs). The correlation between a single indicator and the latency is not strong. Currently, some re-parameterization(Rep) techniques are proposed to convert multi-branch to single-path architecture which is inference-friendly. Nevertheless, multi-branch architectures are still human-defined and inefficient. In this work, we propose a new search space that is suitable for structural re-parameterization techniques. RepNAS, a one-stage NAS approach, is present to efficiently search the optimal diverse branch block(ODBB) for each layer under the branch number constraint. Our experimental results show the searched ODBB can easily surpass the manual diverse branch block(DBB) with efficient training.

Te-Lin Wu, Cheng Li, Mingyang Zhang et al.

Document layout comprises both structural and visual (eg. font-sizes) information that is vital but often ignored by machine learning models. The few existing models which do use layout information only consider textual contents, and overlook the existence of contents in other modalities such as images. Additionally, spatial interactions of presented contents in a layout were never really fully exploited. To bridge this gap, we parse a document into content blocks (eg. text, table, image) and propose a novel layout-aware multimodal hierarchical framework, LAMPreT, to model the blocks and the whole document. Our LAMPreT encodes each block with a multimodal transformer in the lower-level and aggregates the block-level representations and connections utilizing a specifically designed transformer at the higher-level. We design hierarchical pretraining objectives where the lower-level model is trained similarly to multimodal grounding models, and the higher-level model is trained with our proposed novel layout-aware objectives. We evaluate the proposed model on two layout-aware tasks -- text block filling and image suggestion and show the effectiveness of our proposed hierarchical architecture as well as pretraining techniques.

Mingyang Zhang, Xinyi Yu, Jingtao Rong et al.

Automated Machine Learning(Auto-ML) pruning methods aim at searching a pruning strategy automatically to reduce the computational complexity of deep Convolutional Neural Networks(deep CNNs). However, some previous work found that the results of many Auto-ML pruning methods cannot even surpass the results of the uniformly pruning method. In this paper, the ineffectiveness of Auto-ML pruning which is caused by unfull and unfair training of the supernet is shown. A deep supernet suffers from unfull training because it contains too many candidates. To overcome the unfull training, a stage-wise pruning(SWP) method is proposed, which splits a deep supernet into several stage-wise supernets to reduce the candidate number and utilize inplace distillation to supervise the stage training. Besides, A wide supernet is hit by unfair training since the sampling probability of each channel is unequal. Therefore, the fullnet and the tinynet are sampled in each training iteration to ensure each channel can be overtrained. Remarkably, the proxy performance of the subnets trained with SWP is closer to the actual performance than that of most of the previous Auto-ML pruning work. Experiments show that SWP achieves the state-of-the-art on both CIFAR-10 and ImageNet under the mobile setting.

Kun Zhou, Berrak Sisman, Mingyang Zhang et al.

Emotional voice conversion aims to convert the emotion of speech from one state to another while preserving the linguistic content and speaker identity. The prior studies on emotional voice conversion are mostly carried out under the assumption that emotion is speaker-dependent. We consider that there is a common code between speakers for emotional expression in a spoken language, therefore, a speaker-independent mapping between emotional states is possible. In this paper, we propose a speaker-independent emotional voice conversion framework, that can convert anyone's emotion without the need for parallel data. We propose a VAW-GAN based encoder-decoder structure to learn the spectrum and prosody mapping. We perform prosody conversion by using continuous wavelet transform (CWT) to model the temporal dependencies. We also investigate the use of F0 as an additional input to the decoder to improve emotion conversion performance. Experiments show that the proposed speaker-independent framework achieves competitive results for both seen and unseen speakers.

Mingyang Zhang, Xinyi Yu, Jingtao Rong et al.

Previous AutoML pruning works utilized individual layer features to automatically prune filters. We analyze the correlation for two layers from the different blocks which have a short-cut structure. It shows that, in one block, the deeper layer has many redundant filters which can be represented by filters in the former layer. So, it is necessary to take information from other layers into consideration in pruning. In this paper, a novel pruning method, named GraphPruning, is proposed. Any series of the network is viewed as a graph. To automatically aggregate neighboring features for each node, a graph aggregator based on graph convolution networks(GCN) is designed. In the training stage, a PruningNet that is given aggregated node features generates reasonable weights for any size of the sub-network. Subsequently, the best configuration of the Pruned Network is searched by reinforcement learning. Different from previous work, we take the node features from a well-trained graph aggregator instead of the hand-craft features, as the states in reinforcement learning. Compared with other AutoML pruning works, our method has achieved the state-of-the-art under the same conditions on ImageNet-2012.

Andros Tjandra, Berrak Sisman, Mingyang Zhang et al.

We describe our submitted system for the ZeroSpeech Challenge 2019. The current challenge theme addresses the difficulty of constructing a speech synthesizer without any text or phonetic labels and requires a system that can (1) discover subword units in an unsupervised way, and (2) synthesize the speech with a target speaker's voice. Moreover, the system should also balance the discrimination score ABX, the bit-rate compression rate, and the naturalness and the intelligibility of the constructed voice. To tackle these problems and achieve the best trade-off, we utilize a vector quantized variational autoencoder (VQ-VAE) and a multi-scale codebook-to-spectrogram (Code2Spec) inverter trained by mean square error and adversarial loss. The VQ-VAE extracts the speech to a latent space, forces itself to map it into the nearest codebook and produces compressed representation. Next, the inverter generates a magnitude spectrogram to the target voice, given the codebook vectors from VQ-VAE. In our experiments, we also investigated several other clustering algorithms, including K-Means and GMM, and compared them with the VQ-VAE result on ABX scores and bit rates. Our proposed approach significantly improved the intelligibility (in CER), the MOS, and discrimination ABX scores compared to the official ZeroSpeech 2019 baseline or even the topline.

Mingyang Zhang, Xin Wang, Fuming Fang et al.

We investigated the training of a shared model for both text-to-speech (TTS) and voice conversion (VC) tasks. We propose using an extended model architecture of Tacotron, that is a multi-source sequence-to-sequence model with a dual attention mechanism as the shared model for both the TTS and VC tasks. This model can accomplish these two different tasks respectively according to the type of input. An end-to-end speech synthesis task is conducted when the model is given text as the input while a sequence-to-sequence voice conversion task is conducted when it is given the speech of a source speaker as the input. Waveform signals are generated by using WaveNet, which is conditioned by using a predicted mel-spectrogram. We propose jointly training a shared model as a decoder for a target speaker that supports multiple sources. Listening experiments show that our proposed multi-source encoder-decoder model can efficiently achieve both the TTS and VC tasks.